Template Credit: Adapted from a template made available by Dr. Jason Brownlee of Machine Learning Mastery.
SUMMARY: The purpose of this project is to construct a prediction model using various machine learning algorithms and to document the end-to-end steps using a template. The MiniBooNE Particle Identification dataset is a classic binary classification situation where we are trying to predict one of the two possible outcomes.
INTRODUCTION: This dataset is taken from the MiniBooNE experiment and is used to distinguish electron neutrinos (signal) from muon neutrinos (background). The data file is set up as follows. In the first line is the number of signal events followed by the number of background events. The records with the signal events come first, followed by the background events. Each line, after the first line, has the 50 particle ID variables for one event.
In iteration Take1, the script focused on evaluating various machine learning algorithms and identifying the model that produces the best overall metrics. Iteration Take1 established the baseline performance for accuracy and processing time.
For this iteration, we will explore the feature selection technique of eliminating collinear features. By eliminating the collinear features, we hope to decrease the processing time while maintaining an acceptable level of accuracy loss.
ANALYSIS: From the previous iteration Take1, the baseline performance of the eight algorithms achieved an average accuracy of 90.82%. Two algorithms (Bagged CART and Random Forest) achieved the top accuracy scores after the first round of modeling. After a series of tuning trials, Random Forest turned in the top result using the training data. It achieved an average accuracy of 93.74%. By optimizing the tuning parameters, the Random Forest algorithm processed the testing dataset with an accuracy of 93.91%, which was even better than the training data.
In the current iteration Take2, the baseline performance of the machine learning algorithms achieved an average accuracy of 90.04%. Two algorithms (Stochastic Gradient Boosting and Random Forest) achieved the top accuracy metrics after the first round of modeling. After a series of tuning trials, Stochastic Gradient Boosting turned in the top overall result and achieved an accuracy metric of 93.47%. By using the optimized parameters, the Stochastic Gradient Boosting algorithm processed the testing dataset with an accuracy of 93.57%, which was even better than the training data.
From the model-building perspective, the number of attributes decreased by 13, from 50 down to 37. The processing time went from 17 hours 18 minutes in iteration Take1 down to 12 hours 17 minutes in Take2, which was a reduction of 28% from Take1.
CONCLUSION: For this iteration, the Stochastic Gradient Boostin algorithm achieved the best overall results with an improved processing time after eliminating the collinear features. For this dataset, the Stochastic Gradient Boostin algorithm should be considered for further modeling or production use.
Dataset Used: MiniBooNE particle identification Data Set
Dataset ML Model: Binary classification with numerical attributes
Dataset Reference: https://archive.ics.uci.edu/ml/datasets/MiniBooNE+particle+identification
The HTML formatted report can be found here on GitHub.